MOLECULAR AND CELLULAR BIOLOGY, Dec. 1993, p. 7813-7825 Vol. 13, No. 120270-7306/93/127813-13$02.00/0Copyright © 1993, American Society for Microbiology

The Retinoblastoma Protein Binds to a Family of E2FTranscription Factors

JACQUELINE A. LEES,`* MIDORI SAITO,2 MARC VIDAL,1 MARCUS VALENTINE,2THOMAS LOOK,2 ED HARLOW,1 NICHOLAS DYSON,1 AND KRISTIAN HELIN1

Massachusetts General Hospital Cancer Center, Building 149, 13th Street, Charlestown, Massachusetts02129,1 and St. Jude Children's Research Hospital, Memphis, Tennessee 381012

Received 8 August 1993/Returned for modification 18 August 1993/Accepted 15 September 1993

E2F is a transcription factor that helps regulate the expression of a number of genes that are important incell proliferation. Recently, several laboratories have isolated a cDNA clone that encodes an E2F-like protein,known as E2F-1. Subsequent characterization of this protein showed that it had the properties of E2F, but itwas difficult to account for all of the suggested E2F activities through the function of this one protein. Usinglow-stringency hybridization, we have isolated cDNA clones that encode two additional E2F-like proteins,called E2F-2 and E2F-3. The chromosomal locations of the genes for E2F-2 and E2F-3 were mapped to lp36and 6q22, respectfully, confirming their independence from E2F-1. However, the E2F-2 and E2F-3 proteins areclosely related to E2F-1. Both E2F-2 and E2F-3 bound to wild-type but not mutant E2F recognition sites, andthey bound specifically to the retinoblastoma protein in vivo. Finally, E2F-2 and E2F-3 were able to activatetranscription of E2F-responsive genes in a manner that was dependent upon the presence of at least onefunctional E2F binding site. These observations suggest that the E2F activities described previously result fromthe combined action of a family of proteins.

The retinoblastoma gene (RB-1) is one of the best-studiedtumor suppressor genes (reviewed in reference 53). Itscharacterization and cloning were made possible by thefrequent mutation of RB-1 in the development of retinoblas-tomas (15, 16, 34). All retinoblastomas studied to datecontain mutations in both RB-1 alleles, and these mutationslead to the loss or functional inactivation of the retinoblas-toma protein (pRB). Subsequent studies have identified RB-1mutations in a wide variety of other tumors, includingosteosarcomas, small-cell lung carcinomas, breast carcino-mas, prostate carcinomas, and bladder carcinomas (53).Reintroduction of the wild-type RB-I gene into a number ofRB-i-negative cell lines appears sufficient to reverse, or atleast reduce, their tumorigenicity (4, 28, 49, 50, 52). Thesedata suggest that the product of this tumor suppressor genecontributes to the regulation of cellular proliferation in abroad range of tissues.

Genetic studies have shown that an essential portion of thetransforming activity of the small DNA tumor viruses is theability of their oncoproteins (adenovirus ElA, simian virus40 large T antigen, or human papillomavirus E7) to bind andinactivate pRB (reviewed in reference 12). Together, theseobservations suggest that loss of pRB function is one way inwhich abnormal cellular proliferation is frequently induced.

Recent experiments have shown that pRB binds to anumber of cellular proteins and that this binding is disruptedby the viral oncoproteins (reviewed in reference 41). One ofthe best characterized of these pRB-associated proteins isthe transcription factor E2F. E2F was originally identified asa DNA-binding protein that is essential for the ElA-depen-dent activation of the adenovirus E2 promoter (33). How-ever, E2F binding sites have subsequently been found in thepromoters of many cellular genes whose products are impor-tant for cell proliferation (41). Many of these genes areexpressed at the G1/S transition and encode proteins that are

* Corresponding author.

essential for DNA synthesis. However, there are a numberof other E2F-responsive genes that are expressed at otherstages of the cell cycle, including the immediate-early genec-myc (26, 51).The activity of E2F appears to be regulated, at least in

part, by its association with proteins such as pRB. pRBbinds directly to E2F and inhibits its transactivation (19, 21,22, 25, 54). This inhibition of E2F transactivation appears tobe cell cycle regulated. In G1 cells, E2F is associated withunphosphorylated or underphosphorylated pRB (2, 3, 6). Ascells pass the G1/S transition, pRB becomes phosphorylated,and unbound E2F is found. This free E2F is presumed to betranscriptionally active.

In addition to pRB, E2F is also known to interact inde-pendently with the pRB-related protein p107 (5, 9, 46). Thereare two distinct forms of the E2F-p1O7 complex, one con-taining the cell cycle kinase cyclin A/cdk2 and the othercontaining the kinase cyclin E/cdk2 (5, 9, 35, 38). Theformation of these p107-containing complexes is also cellcycle regulated, and their appearance correlates with thetiming of formation of the active kinase complexes (13, 35,46). The role of these p107-containing complexes is less welldefined, but p107 can inhibit E2F-mediated transactivation(44, 58, 59), and the plO7-E2F complexes are also selectivelytargeted by the transforming proteins of the small DNAtumor viruses. These data suggest that pRB, p107, and cellcycle-dependent kinases may act in concert to confine theactivation of E2F and E2F-responsive genes to precisestages of the cell cycle.Using pRB to screen expression libraries, we and others

have recently isolated a cDNA clone, called E2F-1, that hasmany of the properties of E2F (22, 31, 45). E2F-1 binds tounphosphorylated pRB both in vitro and in vivo, and thisbinding can be competed for by adenovirus ElA. It alsointeracts with E2F recognition sites in a sequence-specificmanner. E2F-1 stimulates the transcription of E2F-respon-sive promoters, and this activation can be inhibited by itsdirect association with pRB (21).

7813

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

7814 LEES ET AL.

Despite these findings, a number of experiments suggestedthat E2F-1 is unable to account for all of the observed E2Factivity (11, 22, 31). First, antibodies to E2F-1 are unable tosupershift or abolish all of the E2F DNA binding complexesseen in gel shift experiments. Second, the major E2F-likeproteins associated with pRB and p107 give different proteo-lytic patterns (11). These data suggest that the observed E2Factivity might result from the concerted action of severalE2F-1-related proteins. To test this notion, we have usedlow-stringency hybridization to screen for E2F-1-relatedgenes. Here, we describe the characterization of two pro-teins, designated E2F-2 and E2F-3, which bind to E2F DNArecognition sites, associate with pRB in vivo, and transacti-vate E2F reporter constructs. Similar findings are reportedin the accompanying paper by Ivey-Hoyle and colleagues(30). The isolation of E2F-2 and E2F-3 indicates that theactivity previously described as E2F is actually the com-bined activity of a family of proteins with related properties.

MATERIALS AND METHODS

Isolation of E2F-2 and E2F-3. Initially, 1 million plaquesfrom a human cDNA library, NALM-6 (in ZAPII), werescreened with a fragment of the E2F-1 cDNA, encodingamino acids 89 to 214, that had been labeled with[a-32P]dCTP by random priming (14). Subsequently, a fur-ther 2 million plaques from both a NALM-6 library and ahuman fetal brain library (in ZAPII; Stratagene) werescreened with a mixture of the E2F-1 probe and a randomlyprimed fragment of E2F-2 encoding amino acids 85 to 200.Finally, these filters were rescreened with a randomlyprimed fragment of E2F-3 that encoded amino acids 132 to230. Hybridization was performed at 42°C in 5 x SSC (lxSSC is 0.15 M NaCl plus 0.015 sodium citrate)-5x Den-hardt's solution-30% formamide-5% dextran sulfate-0.5%sodium dodecyl sulfate (SDS)-150 ,ug of sonicated salmonsperm DNA per ml. The filters were washed at roomtemperature three times for 20 min each time in 2x SSC-0.1% SDS. Positive clones were detected by autoradiogra-phy and plaque purified. cDNA clones corresponding toE2F-1 were identified by sequencing with a 20-bp primer,pC15.1 (GTGCTGAAGGTGCAGAAGCG), encoding aminoacids 159 to 165 of E2F-1. By using exonuclease III-S1nuclease digestion to generate nested sets of deletions (24),the remaining novel clones were sequenced with use ofSequenase 2.0 (United States Biochemical Corp.).Genomic DNA cloning. Bacteriophage and cosmid human

genomic DNA libraries were hybridized with 32P-labeledprobes to isolate E2F-1-related genes. These probes in-cluded a 1.0-kb SacII-XhoI fragment of the human E2F-1cDNA, a 0.2-kb KpnI-SmaI genomic DNA fragment of thehuman E2F-2 gene, a 2.2-kb SmaI-SacI fragment of theE2F-2 cDNA, and 0.25-kb EcoRV-HincII and 1.8-kbEcoRV-SmaI fragments of the human E2F-3 cDNA.

Fluorescence in situ hybridization. Bromodeoxyuridine-synchronized, phytohemagglutinin-stimulated peripheralblood lymphocytes from a normal donor were used as asource of metaphase chromosomes. Genomic DNA sub-cloned into phage or cosmid vectors was nick translated withdigoxigenin-11-UTP and hybridized overnight at 37°C tofixed metaphase chromosomes according to the method ofPinkel et al. (42), except for the inclusion of 0.33 ,ug of highlyreiterated human DNA self-annealed to COt = 1 (BethesdaResearch Laboratories, Gaithersburg, Md.) per ,ul. Signalswere detected by incubating the slides with fluorescein-conjugated sheep antidigoxigenin antibodies (Boehringer

Mannheim, Indianapolis, Ind.) followed by counterstainingin propidium iodide solution containing antifade (1,4-diazabicyclo[2.2.2]octane; Sigma Chemical, St. Louis, Mo.).Fluorescence microscopy was performed with a Zeiss stan-dard microscope equipped with fluorescein epifluorescencefilters.

Somatic cell hybrids. A panel of 18 human x mousesomatic cell hybrids containing different human chromo-somes was obtained from the National Institute of GeneralMedical Sciences Human Genetic Mutant Cell Repository(Coriell Institute for Medical Research, Camden, N.J.); thecharacterization and human chromosome content of thesehybrids are described in the repository catalog. We alsoderived seven human x hamster somatic cell lines bypolyethylene glycol-mediated fusion of cryopreserved hu-man leukemic bone marrow blasts with E36 hamster cells, aspreviously described (37). Ten micrograms of DNA fromeach hybrid line was digested with BglII and subjected toSouthern blotting analysis using the 0.2-kb KpnI-SmaIE2F-2 genomic DNA probe, which identified a 7.0-kb humangenomic restriction fragment. The same blots were hybrid-ized with a 0.25-kb EcoRV-HincII E2F-3 cDNA probe,which identifies a 9.5-kb restriction fragment of the humanE2F-3 gene and 8.0- and 6.0-kb fragments of the two E2F-3pseudogenes.

Northern (RNA) blot analysis. Total cellular RNA wasisolated from either ML-1 (premyeloid leukemia), C-33A(cervical carcinoma cell line), U118 (glioblastoma), 293(adenovirus-transformed kidney epithelial cell), or NGP(neuroblastoma) cells by the method of Chomczynski andSacchi (6a). Poly(A)+ RNA was prepared by oligo(dT)-cellulose chromatography (1), and 5 ,ug was fractionated byelectrophoresis on a 1.0% formaldehyde agarose gel andthen transferred to Hybond-N+ membranes (Amersham).The human tissue blot (Clontech Laboratories, Inc.) con-tains 2 ,ug of poly(A)+ RNA from heart, brain, placenta,lung, liver, skeletal muscle, kidney, and pancreas per lane.The blots were screened with fragments of either E2F-1(encoding amino acids 259 to 437), E2F-2 (encoding aminoacids 195 to 437), E2F-3 (encoding amino acids 224 to 425),or human P-actin (as an internal control) that had beenlabeled with [a-32P]dCTP by random priming (14). Hybrid-ization was carried out for 18 h at 42°C in 5x SSPE-lOxDenhardt's solution-50% formamide-2% SDS-100 ,ug ofsonicated salmon sperm DNA per ml. The filters werewashed at room temperature three times for 20 min eachtime in 2x SSC-0.05% SDS and then at 60°C twice for 20min each time in O.lx SSC-0.1% SDS.

Construction of glutathione S-transferase (GST) fusion pro-teins. E2F-2 and E2F-3 sequences were prepared by poly-merase chain reaction with Vent polymerase (New EnglandBioLabs), using the following primers: for E2F-2 (aminoacids 85 to 200), 10.2 and 10.3; for E2F-2 (amino acids 195 to437), 10.11 and 10.17; for E2F-2 (amino acids 244 to 437),10.14 and 10.17; for E2F-2 (amino acids 291 to 437), 10.15and 10.17; for E2F-2 (amino acids 381 to 437), 10.16 and10.17; for E2F-3 (amino acids 132 to 230), 31.2 and 31.3; andfor E2F-3 (amino acids 132 to 270), 31.2 and 31.9.

Sequences of the primers were as follows: 10.2, 5'-GATCCCATGGGATCCCCGGCCAAAAGGAAGCT-3'; 10.3,5'-CTAGGAATTCGTCTTCAAACATTCCCCT-3'; 10.11,5'-CATCCATGGGATCCAGGGGAATGTTTGAAGACCC-3'; 10.14, 5'-GATGGATCCAACAAGAGGCTGGCCTATG-3'; 10.15, 5'-GATGGATCCCTCAAGAGCAGGGAAGG-3';10.16, 5'-GATGGATCCCTCCTGCAGCAGACTGAG-3';10.17, 5'-CTAGAATTCAAGC1TGGCTGTCAGCCTGTCT

MOL. CELL. BIOL.

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

THE E2F PROTEIN FAMILY 7815

GTGA-3'; 31.2,5'-GATCGGATCCATGGCAAAGCGAAGGCTGGA-3'; 31.3, 5'-GATCGGATCCTCAGCCCATCCATTGGACGTTG-3'; and 31.9, 5'-GTACGGATCCTCGGTTAACAGT''GAGGTCC-3'. The amplified fragments were sub-cloned into pGEX2T as BamHI-EcoRI fragments. The result-ing GST fusion proteins were expressed and purified fromEschenchia coli as described previously (48). After binding toglutathione-agarose, the size, purity, and concentration of theGST fusion proteins were evaluated by Coomassie bluestaining of SDS-polyacrylamide gels.Gel retardation assays. Gel retardation assays were per-

formed with 7 to 200 ng of GST fusion in a binding buffercontaining 20 mM N-2-hydroxyethylpiperazine-N'-2-ethane-sulfonic acid (HEPES; pH 7.4), 40 mM KCl, 1mM MgCl2,0.1 mM EDTA, and 0.1% Nonidet P-40 to which 0.1 to 0.5 ngof32P-labeled oligonucleotide probe, corresponding to eitherthe wild-type (5'-ATFLTAAGT'ITCGCGCCC(TTCTCAA-3') or a mutant (5'-ATTTAAGT'lTCGATCCCTTTCTCAA-3') E2F site, was added. The reaction products were sepa-rated on a 4% polyacrylamide gel run at 180 V in 0.25Tris-borate-EDTA at 4°C for 1 h. The gel was then dried andsubjected to autoradiography.

In vitro binding experiments. cRNA was synthesized by invitro transcription using either T7 or T3 RNA polymerase(Promega). The resulting cRNA was translated in rabbitreticulocyte lysate (Promega) in the presence of [35S]methio-nine (New England Nuclear). The synthesized polypeptideswere then incubated for 20 min with 2 ,ug of glutathione-agarose-bound GST fusion protein in ElA lysis buffer(ELB+; 250 mM NaCl, 50 mM HEPES [pH 7.0], 0.1%Nonidet P-40) containing 50 mM NaF, 5 mM EDTA, 0.1 mMsodium orthovanadate, 50,g of phenylmethylsulfonyl fluo-ride per ml, 1 ,ug of leupeptin per ml, 1 ,g of aprotinin perml, 1 mM dithiothreitol. The supernatant was then reincu-bated for 20 min with 2 ,ug of either GST or GST fusionprotein that had been prebound to glutathione-agarose (Sig-ma). The recovered complexes were washed four times withELB+ and then resolved on SDS-10% polyacrylamide gels.

Cell culture and labeling. All cells were cultured at 37°C on100-mm-diameter tissue culture plates in Dulbecco's modi-fied Eagle's medium supplemented with 10% fetal bovineserum. Cells were labeled for 4 h at 37°C with, per plate, 1mCi of Trans[35S]-label (ICN) in 2 ml of methionine-freemedium or 5 mCi of 32Pi (carrier free; ICN) in 1 ml ofphosphate-free medium.

Antisera. Polyclonal antisera to E2F-2 and E2F-3 wereraised by injecting mice with GST fusion proteins containingamino acids 85 to 200 of E2F-2 or amino acids 132 to 230 ofE2F-3 as described by Harlow and Lane (20a). The antibod-ies against pRB (C36 and XZ77), p107 (SD2, SD4, SD6, SD9,and SD15), simian virus 40 large T antigen (PAb419), andE2F-1 have been described previously (11, 20, 22, 27).

Immunoprecipitation-reimmunoprecipitation experiments.Cells were collected by centrifugation and lysed on ice for 30min in 0.5 ml of ELB+. The lysate was precleared with 50 RIof normal rabbit serum and 100 ,u1 of fixed and killedStaphylococcus aureus (Zymed), and the supernatant wasused in immunoprecipitation with either 2 RI of the mousepolyclonal antisera or 100 RI of tissue culture supernatantfrom monoclonal antibody hybridoma cells. The immuno-complexes were recovered on protein A-Sepharose beads,washed four times with ELB+, and then resuspended in 50,ul of 2% SDS-0.1 M dithiothreitol. For reimmunoprecipita-tion experiments, the sample was boiled for 5 min. Thesupematant was then diluted to 1 ml with ELB+ andprecleared with protein A-Sepharose beads prior to incuba-

tion with the second antibody. The proteins were resolvedon SDS-8% polyacrylamide gels and detected by eitherfluorography (35S-labeled proteins) or autoradiography (32P-labeled proteins).

Partial proteolytic mapping. Immune complexes were sep-arated on an SDS-8% polyacrylamide gel. The gel was dried,and the proteins were located by autoradiography. Thebands were excised and digested with various concentra-tions of S. aureus V8 protease as described previously (8).The digestion products were then separated on SDS-20%polyacrylamide gels and visualized by autoradiography.

Transient transfections. The E2F-1 expression constructE2F-1(89-437) has been described previously (22). Expres-sion constructs E2F-2(85-437) and E2F-3(132-425) were pre-pared by polymerase chain reaction with Vent polymeraseby BioLabs using the following primers: for E2F-2(85-437),10.8 (5'-GATCCCATGGGATCCCCGGCCAAAAGGAAGCT-3') and 10.7 (5'-CTAGTCTAGAGGATCCGGCTGTCAGCCTGTCTGTGA-3'); and for E2F-3(132-425), 31.2 (5'-GATCGGATCCATGGCAAAGCGAAGGCTGGA-3') and31.15 (5'-CTAGGATCCGGATCGAAGGAGAGTTCACACGAAGC-3').The amplified fragments were subcloned into pCMV-

neoBam as BamHI fragments, and the coding sequences ofthe resultant expression constructs, pCMVE2F-2(85-437)and pCMVE2F-3(132-425), were confirmed by sequencing.The reporter constructs pE2 wt CAT and pE2 (-64/-60,-45/-36) CAT have been described previously (36). Transfectionswere performed essentially as described previously (18).C-33A cells were grown in 9-cm-diameter tissue culturedishes and transiently transfected with 10,ug of carrier DNA,4,ug of reporter construct, 3p,g of pRSVLUC (54), 0,0.1,0.3,1, or 3p,g of expression plasmid, and pCMV to give a total of20p,g of DNA. Cells were harvested after 24 to 36 h and lysedby freeze-thaw in 200 pl of 25 mM Tris-HCl (pH 8.0), and theextracts assayed for chloramphenicol acetyltransferase andluciferase activities according to standard procedures (10,47).

RESULTS

Isolation of cDNA clones encoding E2F-2 and E2F-3. Todate, all known E2F-responsive genes contain one or moresequences that are closely related to the consensus E2Fbinding site. Therefore, it seemed likely that if other E2F-like proteins existed, they would share considerable homol-ogy with the DNA binding domain of E2F-1. As an initialtest, 1 million recombinant phagie from a human NALM-6cDNA library were screened under low-stringency condi-tions with a radiolabeled probe that encompassed the knownminimal DNA binding domain of E2F-1 (22). Of the 10independent positives identified, 9 continued to hybridize tothe probe under high-stringency conditions, and sequencingconfirmed that they all contained at least some portion of theDNA binding domain of E2F-1. The remaining positiveclone, clone 10, was no longer detected when the filters werewashed under high-stringency conditions. Consistent withthis observation, most of our E2F-1-specific primers failed tohybridize to this clone. However, a single primer (pC15.1),derived from sequences at the C-terminal portion of theDNA binding domain, gave rise to sequence that was relatedbut clearly distinct from E2F-1.

Further sequencing showed that clone 10 contained acontinuous open reading frame of 414 amino acids (Fig. 1A)followed by approximately 3.5 kb of 3' untranslated regionprior to the poly(A)+ tail. This coding sequence is highlyhomologous to that of E2F-1, suggesting that these proteins

VOL. 13, 1993

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

7816 LEES ET AL.

A 1 50

E2f3 .RIQPALE QYLVTAGGGE GAAWAAAAA ASMDNA SPGFAAAAA

51 100E2f2. . PTELNJPSG. Le PATATYYA 3QAPPAAAE2f3 AAAPG&YIQI L1l1NTSTTS. ...C LS GAM..' L SAPGAEIE2fl APAID GALRLLDSSQ VI SAPAPTP A X D..

151 2E2f2 VPE.?T .ID PS;E2f3 YLSDGSP D1tAALRSPE2fl YLAESSGPAREGR.

f201 -250E2f2 Y

Z2f1l V Q DITLw251. 300

E2f2 GD B KPPTSTC; SP $<O STOSME2fl

351 400E2f2 EPLPSTSTPSPSSTSTDPEQNSLAPTVERE2f3 ITPSTD PXPJD ......... LASTE2fl FTP9 Q ..... ..V NRATDSATIVSP

401 455E2f2 ASSVPAPAPT.DI.L.......... D ?E2f NG.... . ..GHL......................... SS

E2fl PPSSPPSSLTD SLSWVDEDR451 500

E2fM OE SPTLA SF"; AZL--04E2f3 klDIS NI...IDGPFV lltII\ G wE2fl SG ...PEEI S .

501 513E2f GDLLIN* ....

E2f EKXLPLVEZDFM CS*E2fl GDLTPLDF.

B89 95 120 191 215 243251 317 409 426

I * H- . SEiCHg,t,5.gy P-,>t,.4............1''

Basic DNA Binding L4ppw! Mbarkd Box pRB Binding

Idinttis 50% 73%/6 28% 55% 56%

Similariis 150%ff 85% 45% 65% 56%

FIG. 1. Sequence comparisons of E2F-1, E2F-2, and E2F-3. (A)Amino acid sequences of E2F-1, E2F-2, and E2F-3. Boxes indicateresidues that are identical to all three proteins. (B) Locations ofconserved motifs in relation to the sequence of E2F-1.

are closely related. Furthermore, the highest degree ofhomology mapped to the region of E2F-1 that is known to beboth necessary and sufficient for DNA binding. In light of theclose homology, we have designated this novel proteinE2F-2.

Despite its large size, clone 10 does not encode an

initiating methionine and must therefore be a partial cDNA.Although we have tried several different strategies, we havebeen unable to extend this clone further. However, duringthe course of this work, we learned that Ivey-Hoyle andcolleagues had also isolated cDNA clones encoding an

E2F-1-related protein (30). After sequence comparisons, itwas clear that these proteins were identical and that we

lacked 29 amino acids of coding sequence. To avoid confu-sion, we have therefore numbered our mutants from theinitiating ATG designated by Ivey-Hoyle et al. (30).

In an attempt to isolate additional E2F family members,we screened 2 million clones from both the NALM-6 library

and a human fetal brain library (Stratagene) with a mixture ofprobes derived for the known DNA binding domain of E2F-1and the putative DNA binding domain of E2F-2. This screengave rise to an additional 52 positive clones, 46 of whichcontained sequences that were identical to E2F-1. Of theremaining clones, five corresponded to the E2F-2 sequence,but all were shorter than the original cDNA clone. Inaddition, this screen identified one additional positive, clone31. This clone contains a 1,275-bp open reading frame that isalmost immediately preceded by an in-frame terminationcodon. By virtue of its homology to E2F-1, the 425-amino-acid protein that is encoded by this clone has been desig-nated E2F-3.To complete this screen, we rehybridized the cDNA

libraries at low stringency with the putative DNA bindingdomain of E2F-3. However, all 36 of the identified positiveclones corresponded to either E2F-1, E2F-2, or E2F-3 itself.It therefore seemed unlikely that this strategy would lead tothe identification of any additional family members.E2F amino acid comparisons. Comparison of the predicted

amino acid sequences of E2F-1, E2F-2, and E2F-3 revealedfour regions of homology (Fig. 1B). Helin et al. (22) havepreviously shown that the DNA binding domain of E2F-1 iscontained within amino acids 89 to 191 of this protein.Comparison with the equivalent region of E2F-2 (aminoacids 85 to 200) and E2F-3 (amino acids 132 to 248) indicatesthat this region contains two highly conserved domains thatare separated by a linker that varies in size from 25 to 32amino acids and maintains little sequence conservation. Thefirst conserved domain corresponds to a small six-amino-acid motif (basic-basic-basic-L-acidic-L) that is highlycharged. The second corresponds to the region of highestsequence identity (73% between all three clones), and itmaps to the carboxy-terminal 71 amino acids of this region.Mapping studies indicate that this carboxy-terminal region iscontained within a single exon (see below and reference 57),and it provides the primary residues required for E2F-1 tocontact DNA (30).The second region of homology is a putative leucine zipper

that spans residues 215 to 243 of E2F-1 (31). Interestingly,there is little sequence conservation in this region apart fromthe hydrophobic residues at the 7 position. Although there isno evidence that this sequence forms an alpha helix, theconservation of this heptad repeat of L-X6-L-X6-L-X6-C-X6-L suggests that this region might serve as a site fordimerization with other proteins or with E2F itself.The third region of homology is in a previously unrecog-

nized area just carboxy terminal to the putative leucinezipper. This region extends from residues 251 to 317 ofE2F-1. Until a function can be ascribed to this region, we arereferring to it as the marked region. The three E2F clonesidentified here show approximately 55% identity in thisregion. When the sequence in the marked region was com-pared with others in the protein data bases, no homologieswere identified. However, this high degree of sequenceconservation suggests that this region may correspond to animportant functional domain.The last segment of homology is in the extreme carboxy

terminus. This region of homology corresponds almost pre-cisely with the region of E2F-1 that was identified as beingboth necessary and sufficient for interaction with pRB. Inthis region, 10 of the 18 residues are identical among E2F-1,E2F-2, and E2F-3.Genomic cloning. Human genomic libraries were screened

with E2F cDNA and genomic probes, and the series ofrecombinant phages and cosmid clones obtained were sub-

MOL. CELL. BIOL.

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

THE E2F PROTEIN FAMILY 7817

2kbA. E2F-2

A29.1

H H BaH H Ba Ba H Ba H. *I.t.I.I, .I,1 4

Bg EBgX E Bg X E Bg E Bg Ea

A10.1

cos 24a

'. cos lOa

B. E2F-313.1

A1.2

H H H H H H4 + + II

E Bg BgX E Bg EBggXb

A14.5

H H H H, I.

BgX X X BgBgEBgEc

D. E2F-3 psoudogene 2H H

+ +

Bg E Bg

EBg

117.1

d

FIG. 2. Genomic restriction maps of the human E2F-2 and E2F-3 loci. Two sets of recombinant bacteriophages and cosmids representingthe authentic E2F-2 (A) and E2F-3 (B) loci are shown, as well as clones from two additional E2F-3 pseudogenes (C and D). Restrictionfragments that hybridized to the cDNA probes are indicated (+) above the maps, and fragments subjected to nucleotide sequencing analysisare demarcated by the open boxes below.

divided into four sets of nonoverlapping sequences, based ontheir restriction maps and preferential hybridization to thehuman E2F cDNAs (Fig. 2A). A fifth locus containing theE2F-1 gene was also isolated and will be reported separately.Two recombinant phages and two cosmids were obtained,and they contained overlapping sequences that hybridized tothe E2F-2 cDNA over a genomic locus spanning approxi-mately 20 kb (Fig. 2A). DNA sequence analysis of a 0.2-kbKpnI-SmaI genomic DNA fragment from this locus (labeled"a" in Fig. 2A) showed that it matched the E2F-2 cDNAsequence within the most conserved, C-terminal portion ofthe DNA binding domain and that it was flanked by consen-sus splice acceptor and donor sequences, confirming thatthis locus encoded E2F-2.Three independent sets of nonoverlapping phages and

cosmids hybridized preferentially under high-stringencyconditions to the human E2F-3 cDNA probes (Fig. 2B to D).Southern blotting analysis indicated that one of the three loci(Fig. 2B) included interrupted coding regions, while theother two loci contained only single hybridizing segments(Fig. 2C and D). Nucleotide sequencing of the fragmentlabeled "b" in Fig. 2B showed that it contained an exonmatching the sequence of the E2F-3 cDNA flanked byconsensus splice donor and acceptor sequences, indicatingthat it encoded the E2F-3 mRNA. By contrast, when hybrid-

izing restriction fragments shown as "c" and "d" in Fig. 2Cand D were sequenced, insertions, deletions, and in-frametermination codons were identified, indicating that these locicontain E2F-3 pseudogenes.Chromosomal assignment ofE2F-2 and E2F-3 genes. E2F is

thought to play an important role in regulating the expressionof genes required for cell proliferation. Therefore, it seemslikely that deregulation of E2F expression or activity maylead to abnormal proliferation. We were interested in iden-tifying the chromosomal localization of E2F-2 and E2F-3 todetermine whether there might be any association withknown diseases.The chromosomal localization of the E2F-2 and E2F-3

genes and the two E2F-3 pseudogenes was determined byfluorescence in situ hybridization, using probes preparedfrom each of the genomic clones hybridized to normalmetaphase human chromosomes (Fig. 3). Upon hybridiza-tion to the cosmid clones containing inserts from the E2F-2gene, the only fluorescence signal emanated from the distalshort arm of chromosome 1. The chromosomal arm wasidentified by cohybridization of the metaphase chromosomeswith pUC1.77, a probe that is specific for chromosome 1heterochromatin (7). On the basis of the distance of thehybridization signal from the centromere relative to theentire length of the short arm of chromosome lp, E2F-2 was

C. E2F-3 pseudogene 1

cos 26a

A7.1

VOL. 13, 1993

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

7818 LEES ET AL.

FIG. 3. Fluorescence in situ nyobriization analysis using digox-igenin-labeled E2F-2 (A) and E2F-3 (B) genomic probes. Humancosmid clones for E2F-2 (cosmids 10a and 24a; Fig. 2A) and E2F-3(cosmid 26a; Fig. 2B) were hybridized to human metaphase chro-mosomes from peripheral blood lymphocytes. Fluorescence signalsresulting from hybridization of these probes are indicated by thewhite arrows at adjacent positions on sister chromatids. Centro-mere-specific probes (arrowheads) for chromosomes 1 (A) and 6 (B)were used to identify the chromosomes containing the E2F-2 andE2F-3 loci.

assigned to band p36. By using the same method, authenticE2F-3 was localized to chromosome band 6p22, based on

cohybridization with the chromosome 6 centromere probeD6Z1 (Oncor, Gaithersburg, Md.). The E2F-3 pseudogenesshown in Fig. 2C and D were localized to chromosome bands17qll-12 and 2q33-35, respectively.These chromosomal assignments were confirmed by hy-

bridizing the E2F-2 and E2F-3 cDNAs and genomic probes

to Southern blots containing DNA extracted from a panel ofrodent x human hybrid cell lines containing different com-binations of human chromosomes (Table 1). A 7.0-kb BglIIrestriction fragment contained within the authentic humanE2F-2 genomic locus was detected concordantly with humanchromosome 1, consistent with the results obtained byfluorescence in situ hybridization. The E2F-3 cDNA probeidentified a 9.5-kb restriction fragment within the humanE2F-3 gene and 6.0- and 8.0-kb fragments within the twoE2F-3 pseudogenes shown in Fig. 2C and D. These frag-ments were identified concordantly with human chromo-somes 6, 17, and 2, again confirming the fluorescence in situhybridization results.

Expression of E2F-2 and E2F-3. Previous studies detectedE2F-1 expression in a wide variety of cell lines and tissues.Therefore, we were interested in comparing the expressionpatterns of E2F-2 and E2F-3 with that of E2F-1. Probes wereprepared by using sequences carboxy terminal of the DNAbinding domain, and their specificity was confirmed by DNAblotting. Northern blot analysis was performed on poly(A)+RNA from ML-1 (premyeloid leukemia), C-33A (a cervicalcarcinoma cell line), U118MG (glioblastoma), 293 (adenovi-rus-transformed kidney epithelial cell), and NGP (neuroblas-toma) cells (Fig. 4A). Consistent with the sizes of the cDNAclones, both E2F-2 and E2F-3 give rise to large transcripts of5.5 to 6 kb. These mRNAs are present at considerably lowerlevels than that of E2F-1 but have been detected in all of thecell lines examined to date. Interestingly, although the totallevel of expression varies considerably from one cell line tothe next (even accounting for the differences in actin levels),the relative ratio of E2F-1 to E2F-2 to E2F-3 remainsconstant.

In independent tests, we also examined the expression ofthese transcripts in U20S and SAOS-2 (two osteosarcomacell lines), UOC-B1 (a pro-B-cell lymphoblastic leukemiacell line), HEL (an erythroleukemia cell line), NB1 and NB2(two neuroblastoma cell lines), Rhl8 (a rhabdomyosarcomacell line), Colo32ODM (a colonic adenocarcinoma cell line),and SW900 (a squamous cell lung carcinoma cell line). Allshowed similar-size transcripts, as shown in Fig. 4A.

Interestingly, when RNA gels were run for longer timesprior to blotting, the E2F-3 band consistently resolved into adoublet while the E2F-1 and E2F-2 specific probes continuedto detect a single band (data not shown). The significance ofthese two messages is unclear.We have also examined the expression of these genes in

normal human tissues: heart, brain, placenta, lung, liver,skeletal muscle, kidney, and pancreas (Fig. 4B). E2F-2mRNA levels were highest in placenta and then lung andkidney. Lower levels were seen in liver and pancreas. Inskeletal muscle, the 6.0-kb transcript was absent, but anapproximately 2.5-kb mRNA was easily detected. We do notknow the origin of this smaller RNA species. E2F-3 appearsto be expressed in most of these tissues except brain atroughly equivalent levels (compared with the levels of actincontrols). The E2F-3 doublet seen in tissue culture cell lineswas not apparent in tissue Northern analysis.DNA binding activity. If these novel genes are genuine

members of the E2F family of transcription factors, wewould expect them to bind to consensus E2F sites in amanner similar to that of E2F-1. The minimal DNA bindingdomain of E2F-1 has been mapped to amino acids 89 to 191(22). Therefore we tested the ability of the equivalent regionsof E2F-2 (amino acids 85 to 200) and E2F-3 (amino acids 132to 230) to bind DNA as GST fusion proteins. In theseexperiments, E2F-2 was able to bind to DNA in a manner

MOL. CELL. BIOL.

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

V,THE E2F PROTEIN FAMILY 7819

TABLE 1. Assignment of E2F-2 and E2F-3 genes to human chromosome in rodent x human somatic cell hybrids

Assignment of E2F-2 Assignment of E2F-3HumancHromosome No. of hybrids No. of hybridschromosome o ,o

+/+ +1- -1+ Discordancy +/+ Discordancy

1 3 0 0 22 0 1 6 2 16 322 1 2 2 20 16 3 4 0 18 163 1 2 7 15 36 5 2 3 15 204 1 2 0 23 8 1 6 0 18 245 0 3 3 19 24 2 5 1 17 246 1 2 6 16 28 7 0 0 18 07 1 2 3 19 20 4 3 0 18 128 2 1 3 19 16 4 3 1 17 169 0 3 3 19 24 0 7 3 15 4010 1 2 1 21 12 2 5 0 18 2011 0 3 3 19 24 2 5 2 16 2812 0 3 7 15 40 5 2 2 16 1613 2 1 3 19 16 3 4 2 16 2414 1 2 2 20 16 1 6 2 16 3215 2 1 2 20 12 3 4 1 17 2016 1 2 2 20 16 3 4 0 18 1617 1 2 2 20 16 2 5 1 17 2418 1 2 5 17 28 5 2 1 17 1219 1 2 2 20 16 2 5 1 17 2420 2 1 3 19 16 3 4 2 16 2421 1 2 6 16 32 5 2 2 16 1622 0 3 6 16 36 4 3 2 16 20X 2 1 10 12 44 5 2 7 11 36Y 0 3 1 21 16 0 7 1 17 32

that was almost indistinguishable from that of E2F-1 (Fig.5A). In contrast, this region of E2F-3 displayed little or noDNA binding activity in these experiments (data not shown).However, a slightly larger region of this protein, encompass-ing amino acids 132 to 289, did display sequence-specificDNA binding activity (Fig. 5A). However, when we tested arange of input fusion protein concentrations, it was clear thatthe GST-E2F-3 fusion protein bound to DNA with signifi-cantly reduced affinity or avidity relative to E2F-1 and E2F-2(Fig. SB). We are currently trying to determine whether thisreduced binding reflects genuine differences in the affinityand/or sequence specificity of this protein or whether itmerely reflects the ability of these GST-fusion proteins toadopt the correct conformation in vitro.

Recently, Huber and coworkers demonstrated that E2Fbinds to DNA as a heterodimer (29). However, mixingexperiments failed to reveal any synergy between E2F-1,E2F-2, and E2F-3 in gel retardation assays (data not shown).E2F-2 and E2F-3 bind to pRB in vivo. The pRB binding

domain has been mapped to an 18-amino-acid motif at thecarboxy terminus of E2F-1 (22). Since this motif is highlyconserved in both E2F-2 and E2F-3, these proteins alsoseemed likely to bind either pRB or the related protein p107.To test the binding properties of these proteins, we raisedpolyclonal antisera that are specific for E2F-2 or E2F-3 andused them (alongside an E2F-1 antiserum [22]) in an immu-noprecipitation-reimmunoprecipitation assay to determinewhether E2F-2 or E2F-3 could be detected in associationwith pRB or p107 in vivo. ML-1 cells were labeled with[35S]methionine and then immunoprecipitated with monoclo-nal antibodies specific for pRB (C36 and XZ77) or p107(SD2, SD4, SD6, SD9, and SD15). The samples were thendenatured by heating in 2% SDS and reimmunoprecipitatedwith either E2F-1-, E2F-2-, or E2F-3-specific polyclonalantisera or an antipeptide antibody that was raised against

the 18-amino-acid pRB binding domain of E2F-1 (Fig. 6A).As previously described (11, 22), the E2F-1-specific poly-clonal antiserum recognized a single protein species that ispresent in pRB but not p107 immunoprecipitations. In con-trast, the anti-E2F peptide antibody recognized at least twoproteins in the pRB immunoprecipitate, the lower one ofwhich comigrates exactly with the E2F-1-specific band. Inaddition, this antiserum detects several bands within thep107 immunoprecipitate that appear to be distinct from thepRB-associated proteins recognized by this antiserum (11).The anti-E2F-2 polyclonal antiserum failed to recognize

any proteins from the p107 immunoprecipitation (Fig. 6A).However, this antiserum clearly detected a single, majorband in the pRB immunoprecipitations. Interestingly, thisE2F-2 species comigrated exactly with the larger proteinspecies recognized by the anti-E2F peptide antibody. Simi-larly, the anti-E2F-3 antibodies immunoprecipitated twoproteins that also appear to be specifically associated withpRB but not p107. We are currently trying to determine theexact identities of these two species.The comigration observed in this experiment suggested

that the pRB-associated proteins detected by the anti-E2Fpeptide antibody correspond to E2F-1 and E2F-2. Afterseveral attempts, it became clear that levels of these 35S-labeled species were too low for us to be able to use V8partial proteolytic mapping to confirm this. However, when2pi was used to label cells, we continued to detect the samepattern of pRB- and p107-associated proteins (Fig. 6B), butthe labeling was increased to sufficient levels to make V8partial proteolytic mapping feasible. ML-1 cells were there-fore labeled with 32Pi, and lysates were immunoprecipitatedwith monoclonal antibodies to pRB. The samples were thendenatured by heating in 2% SDS and reimmunoprecipitatedwith either an E2F-1-, E2F-2-, or E2F-3-specific polyclonalantiserum or the anti-E2F peptide antibody and separated on

VOL. 13, 1993

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

7820 LEES ET AL.

C?n c, Z-

nOI? 0

mn~ C,)

2 no0 z2 n o(. C\j

28S -

18S -

E2F-2

.. ..#.

; .. .f.... .. * . .*F_

h.. 'wF'

3 Actin

V)

E a) >.

M , nCCD CICCL 3

9.5 _7.5 -

4.4 -

2.4 -

1.35 -

E2F-2

InC (D >, Q

=5 '{1CD

a: cn Cs c

I3 C COIAL >X

'aY C_rccaij-L2 E

4:.,* *iWI......

a'

E2F-3

FIG. 4. Expression of E2F-2 and E2F-3 mRNAs, determined by Northern blot analysis of poly(A)+ RNA from tissue culture cell lines (A)or normal human tissues (B) using probes specific for either E2F-1, E2F-2, or E2F-3. Dashes indicate the positions of 28S and 18S rRNAsubunits (tissue culture cell lines) or RNA markers (human tissues). Exposure times for the cell lines/tissue blots were as follows: E2F-1, 2days/0; E2F-2, 9 days/14 days; E2F-3, 9 days/14 days; and actin 1 h/3 h.

SDS-8% polyacrylamide gels (Fig. 6C). (The smaller speciesin the E2F-3 immunoprecipitate is absent in this experimentbecause it has been eluted from the bottom of the gel.) Eachof the bands was then subjected to V8 partial proteolyticmapping (Fig. 6D). E2F-1, E2F-2, and E2F-3 each gave riseto a distinct pattern of partial digestion fragments. More-over, the partial proteolytic maps of the two bands recog-nized by the anti-E2F peptide antibody were identical tothose of E2F-1 (lower band) and E2F-2 (upper band), respec-tively. Despite the high degree of homology within the pRBbinding domain of these three proteins, E2F-3 is not recog-nized by the antipeptide antibody, and its V8 partial proteo-lytic map is clearly distinct from those of the two speciesrecognized by this antibody.Because we were unable to detect any p107 binding in this

cell line, we repeated this immunoprecipitation-reimmuno-precipitation experiment by using a cervical carcinoma cellline, C-33A, that lacks functional pRB protein. Even in theabsence of pRB, we failed to detect any association betweenp107 and E2F-1, E2F-2, or E2F-3 in these assays (data notshown). Although we cannot rule out the possibility thatthese proteins show a low level of binding to p107, these datasuggest that E2F-1, E2F-2, and E2F-3 bind preferentially topRB in vivo.

The pRB binding domain maps to the C-terminal portion.To further map the pRB binding domain, the carboxy-terminal 243 [E2F-2(195-437)], 194 [E2F-2(244-437)], 146[E2F-2(291-437)], or 56 [E2F-2(381-437)] amino acids ofE2F-2 were expressed as GST fusion proteins and tested forthe ability to bind to in vitro-translated pRB, p107, orluciferase as a control. Levels of binding were also com-pared with those for both GST alone and GST-E1A (Fig. 7).As expected, GST-E1A bound with approximately equalaffinity to both pRB and p107, with binding limited only tothe larger polypeptides that are known to maintain theintegrity of the ElA-binding pocket. Consistent with the invivo binding results, we were able to detect little or nobinding of any of the E2F-2 mutants to p107. In contrast, allof the mutants displayed high-affinity binding to pRB, in amanner that was almost indistinguishable from that of ElA.Although such in vitro binding experiments do not prove theidentity and/or specificity of potential binding partners, thesedata do support the strong preference for pRB bindingobserved in vivo. Furthermore, all of the E2F-2 deletionmutants bound specifically to pRB, indicating that the pRBbinding domain lies at the C-terminal end of the protein, aspreviously described for E2F-1 (22).We also tested a similar series of E2F-3 deletion mutants

E2F-1 E2F-2

BU)

Acti n

MOL. CELL. BIOL.

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

THE E2F PROTEIN FAMILY 7821

FreGST

Free E2F-1

(89-191)

GSTE2F-2

(85-200)

GSTE2F-3

(132-270)GST

I I I 11 11] 11 1~

E ,E '5 E E,E5 - E,

b ~ ~~~~~~~*- i

_.

(Fig. 8A) or E2F-3 (Fig. 8B) reproducibly resulted in a 5- to15-fold activation of the wild-type E2 promoter. This activa-tion was independent of the presence of a functional ATFsite (data not shown) but was completely abolished in theabsence of the two E2F sites (Fig. 8). Although theseinductions appear relatively small, they are increased by upto 50-fold when the reporter contains multiple copies of theconsensus E2F site (data not shown). Together, these exper-iments demonstrate that E2F-2 and E2F-3 are both capableof activating transcription in a manner that is dependentupon the presence of at least one functional E2F site.

B GST-E2F-1 GST-E2F-2(89-191) (85-200)

CD c CD cm C CDInput levels- CDO O 5 0 0 0 0 c

N- N- N- c N- r-1N- N- C' N-

Competitor- ! 1

GST-E2F-3(132-270)

la) a) a CC c) c

o o Cco o o 0 0

cm c( cm N- c

FIG. 5. E2F-2 and E2F-3 bind to E2F DNA recognition se-quences. (A) Ten micrograms of 293 cell extract (free E2F), 70 ng ofpurified GST-E2F-1 (amino acids 89 to 191) or GST-E2F-2 (aminoacids 85 to 200), or 200 ng of GST-E2F-3 (amino acids 132 to 270) orfree GST was incubated with a labeled oligonucleotide containing asingle E2F binding site. Wild-type (wt) or mutant (mut) competitoroligonucleotides were added at a 100-fold excess as indicated. (B)Different input levels of GST-E2F-1 (70 to 7 ng), GST-E2F-2 (70 to7 ng), or GST-E2F-3 (200 to 20 ng) were incubated with the E2Foligonucleotide probe in the presence or absence of competitoroligonucleotides as indicated.

in this assay. Surprisingly, these mutants failed to bind eitherpRB or p107 (data not shown). Since E2F-3 is clearlyassociated with pRB and not p107 in vivo, we assume thatthis lack of binding reflects either aberrant folding or a strongrequirement for posttranslational modifications.

Transactivation. Finally, we compared the ability of theseE2F family members to activate transcription from either thewild-type E2 promoter ([+E2F]) or mutant E2 promoters inwhich either the ATF (-80 to -70) or the two E2F (-64 to-60 and -45 to -36) ([-E2F]) sites have been mutated (36).These reporters were transiently transfected into C-33A, ahuman cervical carcinomas cell line that expresses a func-tionally inactive form of the pRB protein, along with increas-ing levels of an E2F-1 [pCMVE2F-1(89-437)], E2F-2[pCMVE2F-2(85-437)], or E2F-3 [pCMVE2F-3(132-425)] ex-pression plasmid. In these experiments, expression of E2F-2

DISCUSSION

E2F is a cellular transcription factor that appears to play amajor role in regulating the expression of genes that areessential for cell cycle progression. By virtue of its associa-tion with pRB, we and others have previously cloned aprotein, called E2F-1, that displays properties of the endo-genous E2F. Using low-stringency hybridization, we haveisolated cDNA clones that encode two additional E2F-likeproteins, E2F-2 and E2F-3. These three proteins appear tobe closely related, with the highest degree of identity map-ping to regions of the E2F-1 protein that are known to berequired for either DNA binding, pRB binding, or transcrip-tional activation.These novel proteins share many of the biochemical

properties of E2F-1. First, E2F-1, E2F-2, and E2F-3 all bindto wild-type but not mutant E2F sites. Second, when trans-fected into tissue culture cells, E2F-1, E2F-2, and E2F-3 areall capable of activating transcription of known E2F-respon-sive genes in a manner that is dependent upon the presenceof one or more intact E2F sites. Finally, using polyclonalantisera, we can detect association between pRB and each ofthese proteins in vivo. In contrast, we have been unable todetect any association between E2F-1, E2F-2, or E2F-3 withp107 in these assays. Although we cannot rule out thepossibility that a fraction of these proteins associate withp107, we believe that these three proteins represent onedistinct subclass of E2F, for which pRB is the major regu-lator.Using peptide sequence from purified protein prepara-

tions, Girling and colleagues recently isolated a cDNA cloneencoding one component of the murine equivalent of E2F(17). They designated this clone DP-1. DP-1 bears somehomology to E2F-1, E2F-2, and E2F-3, limited primarily toa domain that corresponds to the carboxy-terminal portionof their minimal DNA binding domains. DP-1 has beenshown to bind specifically to consensus E2F sites (17),indicating that this protein is also a member of this family oftranscription factors. Recently, Helin et al. have demon-strated that E2F-1 and DP-1 can interact to generate stableheterodimers (23). Moreover, this interaction leads to anincrease in the DNA binding, pRB binding, and activation ofE2F-responsive transcription. Since E2F-2 and E2F-3 sharemany of the properties of E2F-1, it seems likely that theseproteins will also function as heterodimers in vivo. Interest-ingly, Wu and Harlow have recently isolated at least oneadditional DP-1-like protein, suggesting that there is also afamily of DP proteins (56).

Together, these data suggest that the endogenous E2Factivity results from the concerted activity of multiple het-erodimers that contain one E2F-like and one DP-like com-ponent. Similar multiplicity of heterodimeric subunits hasbeen well documented for a number of other transcription

A

VOL. 13, 1993

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

7822 LEES ET AL.

A pRB

MOL. CELL. BIOL.

p107 B Control pRB pl07

aa0K- N Ct)

cm aL 04 C') aLL U-

L ILLIL0 N 1A C4U. U- LCY U-C :U- U- NL

;: UJ NUJ 52 CM CMw i

c c c c aO C C CO C C C C

.Uw i

.un

....

:.

0-

CD CD CDa a a

L L IL- U- ULL LL C LL L LLCMcl C\iN\CQ1 \NC 1 N O1 N 11U40LUMz<<< <.<O CC O CEE OcCCC E C

O < C <CoCCc66< O)C:CC'c<

D 0a)

0 c O)caco 04

LU Lu a) JO

UC- CL CN

'C C C C< < < < cl

LO 0) 0) LO 0 LO) 00C 0 0 LO 0) 0

oD Lo 0 0o 0 0 U)o0LO 0 0 LO

6 6L6 6 6 6 6 6t6 661ui666)

oooO

FIG. 6. E2F-2 and E2F-3 bind to pRB in vivo. (A to C) ML-1 cells were labeled with either [35S]methionine (A) or 32p; (B and C). The lysates

were immunoprecipitated with monoclonal antibodies specific for either pRB (XZ77 and C36) or p107 (SD2, SD4, SD6, SD9, and SD15) as

indicated. These precipitates were then boiled in 2% SDS-0.1 M dithiothreitol, and the supematant was diluted for immunoprecipitation witheither a polyclonal antiserum raised against GST-E2F-1, GST-E2F-2, or GST-E2F-3 fusion protein or an antipeptide antibody (anti-E2Fpeptide) raised against the 18-amino-acid minimal pRB binding domain of E2F-1. PAb419 (first immunoprecipitation) and a mixture of normalmouse and normal rabbit sera (reimmunoprecipitation) were used as negative controls in these experiment. (D) The major proteins species wereexcised form the gel shown in panel C, subjected to partial V8 proteolytic digestion, and separated on SDS-20% polyacrylamide gels.

C

,:

ii,

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

THE E2F PROTEFTN FAMTLY 7823

GST-E2F-2CDCu I 'I0) 4

0) st 0)7 COCO 04MC: c

GST-E2F-2

I Im , ,~~~~<r:

a: M cn Xn Xn Lua I 1t 14 It _L QL F02 cnen >

GST-E2F-2

1 <~~~~cN.N.NN

"I t-_0z) St 0 a0

040C4CD CD

200 K-

92.5 K-

a"W.

69 K-

46 K j

30 K-

w

-a

mA-si

Luciferase pRB pl107FIG. 7. The pRB binding domain maps to the C-terminal portion of E2F-2. Twenty microliters of in vitro-translated (IVT) luciferase, pRB

(amino acids 1 to 928), or p107 (amino acids 190 to 1068) was incubated with 2 pug of GST, GST-E2F-2 (amino acids 195 to 437), E2F-2 (aminoacids 244 to 437), E2F-2 (amino acids 291 to 437), E2F-2 (amino acids 381 to 437), or GST-13S ElA as indicated. Two microliters of each totalin vitro translation reaction was included on the gel for comparison.

factors, including the families of Fos and Jun proteins (32,40, 43) and the basic/helix-loop-helix proteins (39).The majority of cell lines that we have tested appear to

express E2F-1, E2F-2, and E2F-3. Why do cells requirethree independent pRB-regulated E2Fs? It is possible that

these proteins are functionally redundant and that the cellmaintains multiple genes to ensure that the regulation of thiskey cell cycle regulator cannot be easily abrogated. Alterna-tively, E2F-1, E2F-2, and E2F-3 may have subtly differentfunctions that give rise to distinct elements of E2F activity.

I~-

F~-04c

0

-a

4cC.

0 1 2 3INPUT EXPRESSION VECTOR (ug)

- E2F-2 [+E2F]

----a--- E2F-1 [+E2F]

I~-

4c

-i

4C.

---4--- E2F-1 [-E2F]0 E2F-2 [-E2F]

0 1 2 3

----C--- E2F-1 [+E2F]E2F-3 [+E2F]

E2F-1 [-E2F]E2F-3 [-E2FJ

INPUT EXPRESSION VECTOR (ug)

FIG. 8. E2F-2 and E2F-3 transactivate E2F reporter constructs. E2F-2 (amino acids 85 to 437) (A) and E2F-3 (amino acids 132 to 425) (B)were tested in transient transfection assays for the ability to activate transcription from reporter constructs containing either the wild-type([+E2F]) or a mutant ([-E2F]) site. C-33A cells were transfected in duplicate with the indicated amounts of expression plasmidspCMVE2F-1(89-437), pCMVE2F-2(85-437), and pCMVE2F-3(132-425) along with 4 ,g of either pE2 wt CAT [+E2F] or pE2 (-64/-60,-45/-36) CAT [-E2F], 3 ,g of pRSVLUC, and carrier to give a total of 20 p,g. CAT, chloramphenicol acetyltransferase; LUC, luciferase.

VOL. 13, 1993

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

7824 LEES ET AL.

Although there is no strong evidence to support this hypoth-esis, it is entirely consistent with the emerging view that theregulation of E2F is a highly complex process. In particular,we have to account for the observation that there aremultiple regulators of E2F activity that are active at distinctbut overlapping stages of the cell cycle and that the con-certed activity of these proteins appears to be sufficient togive rise to differences in both the degree and timing ofactivation of the known responsive promoters.The existence of multiple E2F-1-like genes may contribute

to this complex regulation in several ways. First, theseproteins may have some differences in their DNA binding,pRB binding, or transcriptional activities that are sufficientto account for the differential expression of different subsetsof genes. Indeed, our gel retardation experiments suggestthat E2F-3 binds to consensus E2F sites with considerablylower affinity than E2F-1 or E2F-2. Similarly, considerablyhigher levels of E2F-2 and E2F-3 are required to ensure themaximal activation of the E2F-responsive reporters. Inaddition, the properties of the heterodimeric partner arelikely to make a significant contribution to these activities.We are currently conducting a systematic analysis of eachE2F-1-related protein, in combination with each DP-1-re-lated protein, to try to establish the correct partners for eachof these proteins.Two additional levels of control are also likely to exist.

First, the activities of E2F-1, E2F-2, and E2F-3 may bedifferentially regulated throughout the cell cycle. Second, wehave noted that there are differences in the tissue distribu-tion of E2F-1, E2F-2, and E2F-3. For example, we wereunable to detect E2F-2 and E2F-3 mRNAs in brain, thetissue that contains the highest level of E2F-1 expression. Apotential consequence of this observation is that differentgenes may be E2F responsive in different cells. In contrast,cells containing multiple E2Fs may use permutations ofdifferent heterodimers to give rise to graded responses todifferent combinations of input signals. Further investigationof this family of proteins will be essential if we are tounderstand fully the mechanism by which E2F-dependenttranscription is regulated.

ACKNOWLEDGMENTS

We thank Brian Dynlacht for careful reading of the manuscript.J.A.L. was a fellow of the Human Frontiers Science Organization

during a portion of this work. M.V. is supported by the BelgiumNational Fund for Scientific Research. K.H. is a fellow of theDanish Medical Research Council. E.H. is an American CancerSociety Research Professor. This work was supported by NIHgrants to E.H. and T.L.

Midori Saito and Marc Vidal contributed equally to this work.

REFERENCES1. Aviv, H., and P. Leder. 1972. Purification of biologically active

globin messenger RNA by chromatography on oligothymidylicacid-cellulose. Proc. Natl. Acad. Sci. USA 69:1408-1412.

2. Bagchi, S., R. Weinmann, and P. Raychaudhuri. 1991. Theretinoblastoma protein copurifies with E2F-I, an ElA-regulatedinhibitor of the transcription factor E2F. Cell 65:1063-1072.

3. Bandara, L. R., and N. B. La Thangue. 1991. Adenovirus Elaprevents the retinoblastoma gene product from complexing witha cellular transcription factor. Nature (London) 351:494-497.

4. Bookstein, R., J. Y. Shew, P. L. Chen, P. Scully, and W. H. Lee.1990. Suppression of tumorigenicity of human prostate carci-noma cells by replacing a mutated RB gene. Science 247:712-715.

5. Cao, L., B. Faha, M. Dembsld, L.-H. Tsai, E. Harlow, and N.Dyson. 1992. Independent binding of the retinoblastoma protein

and p107 to the transcription factor E2F. Nature (London)355:176-179.

6. Chellappan, S., S. Hiebert, M. Mudryj, J. Horowitz, and J.Nevins. 1991. The E2F transcription factor is a cellular target forthe RB protein. Cell 65:1053-1061.

6a.Chomczynski, P., and N. Sacchi. 1987. Single-step method ofRNA isolation by acid guanidinium thiocyanate-phenol-chloro-form extraction. Anal. Biochem. 162:156-159.

7. Christiansen, H., J. Schestag, W. Bielke, B. Schutz, G. Rust, R.Engel, E. Beniers, N. Christiansen, and F. Lampert. 1991.Chromosome 1 interphase-cytogenetics in 32 primary neuro-blastomas of different clinical stages. Adv. Neurol. 3:99-105.

8. Cleveland, D. W., S. G. Fischer, M. W. Kirschner, and U. K.Laemmli. 1977. Peptide mapping by limited proteolysis in so-dium dodecyl sulfate and analysis by gel electrophoresis. J.Biol. Chem. 252:1102-1106.

9. Devoto, S. H., M. Mudryj, J. Pines, T. Hunter, and J. R. Nevins.1992. A cyclin A-specific protein kinase complex possessessequence-specific DNA binding activity: p33cdk2 is a compo-nent of the E2F-cyclin A complex. Cell 68:167-176.

10. De Wet, J. R., K. V. Wood, M. DeLuca, D. R. Helinski, and S.Subramani. 1987. Firefly luciferase gene: structure and expres-sion in mammalian cells. Mol. Cell. Biol. 7:725-737.

11. Dyson, N., M. Dembski, A. Fattaey, C. Ngwu, M. Ewen, and K.Helin. 1993. Analysis of p107-associated proteins: p107 associ-ates with a form of E2F that differs from pRB-associated E2F-1.J. Virol. 67:7641-7647.

12. Dyson, N., and E. Harlow. 1992. Adenovirus ElA targets keyregulators of cell proliferation. Cancer Surv. 12:161-196.

13. Faha, B., E. Harlow, and E. Lees. 1993. The adenovirus ElA-associated kinase consists of cyclin E-p33cdk2 and cyclinA-p33cdk2. J. Virol. 67:2456-2465.

14. Feinberg, A. P., and B. Vogelstein. 1983. A technique forradiolabeling DNA restriction endonuclease fragments to highspecific activity. Anal. Biochem. 132:6-13.

15. Friend, S. H., R. Bernards, S. Rogelj, R. A. Weinberg, J. M.Rapaport, D. M. Albert, and T. P. Dryja. 1986. A human DNAsegment with properties of the gene that predisposes to retino-blastoma and osteosarcoma. Nature (London) 323:643-646.

16. Fung, Y. K., A. L. Murphree, A. T'Ang, J. Qian, S. H. Hinrichs,and W. F. Benedict. 1987. Structural evidence for the authen-ticity of the human retinoblastoma gene. Science 236:1657-1661.

17. Girling, R., J. F. Partridge, L. R. Bandara, N. Burden, N. F.Totty, J. J. Hsuan, and N. B. La Thangue. 1993. A newcomponent of the transcription factor DRTF1/E2F. Nature(London) 362:83-87.

18. Graham, F. L., and A. J. van der Eb. 1973. A new technique forthe assay of infectivity of human adenovirus. Virology 52:456-467.

19. Hamel, P. A., R. M. Gill, R. A. Phillips, and B. L. Gallie. 1992.Transcriptional repression of the E2-containing promotersEIIaE, c-myc, and RB1 by the product of the RB1 gene. Mol.Cell. Biol. 12:3431-3438.

20. Harlow, E., L. V. Crawford, D. C. Pim, and N. M. Williamson.1981. Monoclonal antibodies specific for simian virus 40 tumorantigens. J. Virol. 39:861-869.

20a.Harlow, E., and D. Lane. 1988. Antibodies: a laboratory man-ual. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y.

21. Helin, K., E. Harlow, and A. R. Fattaey. 1993. Inhibition ofE2F-1 transactivation by direct binding of the retinoblastomaprotein. Mol. Cell. Biol. 13:6501-6508.

22. Helin, K., J. A. Lees, M. Vidal, N. Dyson, E. Harlow, and A.Fattaey. 1992. A cDNA encoding a pRB-binding protein withproperties of the transcription factor E2F. Cell 70:337-350.

23. Helin, K., C.-L. Wu, A. R. Fattaey, J. A. Lees, B. D. Dynlacht,C. Ngwu, and E. Harlow. 1993. Heterodimerization of thetranscription factors E2F-1 and DP-1 leads to cooperativetransactivation. Genes Dev. 7:1850-1861.

24. Henikoff, S. 1984. Unidirectional digestion with exonuclease IIIcreates targeted breakpoints for DNA sequencing. Gene 28:351-359.

25. Hiebert, S. W., S. P. Chellappan, J. M. Horowitz, and J. R.

MOL. CELL. BIOL.

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.

THE E2F PROTEIN FAMILY 7825

Nevins. 1992. The interaction of pRb with E2F inhibits thetranscriptional activity of E2F. Genes Dev. 6:177-185.

26. Hiebert, S. W., M. Lipp, and J. R. Nevins. 1989. ElA-dependenttrans-activation of the human MYC promoter is mediated by theE2F factor. Proc. Natl. Acad. Sci. USA 86:3594-3598.

27. Hu, Q., C. Bautista, G. Edwards, D. Defeo-Jones, R. Jones, andE. Harlow. 1991. Antibodies specific for the human retinoblas-toma protein identify a family of related polypeptides. Mol.Cell. Biol. 11:5792-5799.

28. Huang, H. J., J. K. Yee, J. Y. Shew, P. L. Chen, R. Bookstein,T. Friedmann, E. Y. Lee, and W. H. Lee. 1988. Suppression ofthe neoplastic phenotype by replacement of the RB gene inhuman cancer cells. Science 242:1563-1566.

29. Huber, H. E., G. Edwards, P. Goodhart, D. R. Patrick, P. S.Huang, M. Ivey-Hoyle, S. F. Barnett, A. Oliff, and D. C.HeimbrooL 1993. Transcription factor E2F binds DNA as aheterodimer. Proc. Natl. Acad. Sci. USA 90:3525-3529.

30. Ivey-Hoyle, M., R. Conroy, H. E. Huber, P. J. Goodhart, A.Oliff, and D. C. HeimbrooL 1993. Cloning and characterizationof E2F-2, a novel protein with the biochemical properties oftranscription factor E2F. Mol. Cell. Biol. 13:7802-7812.

31. Kaelin, W. G., W. Krek, W. R. Sellers, J. A. DeCaprio, F.Ajchanbaum, C. S. Fuchs, T. Chittenden, Y. Li, P. J. Farnham,M. A. Blanar, D. M. Livingston, and E. K. Flemington. 1992.Expression cloning of a cDNA encoding a retinoblastoma-binding protein with E2F-like properties. Cell 70:351-364.

32. Kouzarides, T., and E. Ziff. 1988. The role of the leucine zipperin the Fos-Jun interaction. Nature (London) 336:646-651.

33. Kovesdi, I., R. Reichel, and J. R. Nevins. 1986. Identification ofa cellular transcription factor involved in ElA trans-activation.Cell 45:219-228.

34. Lee, W. H., R. Bookstein, F. Hong, L. J. Young, J. Y. Shew, andE. Y. Lee. 1987. Human retinoblastoma susceptibility gene:cloning, identification, and sequence. Science 235:1394-1399.

35. Lees, E., B. Faha, V. Dulic, S. I. Reed, and E. Harlow. 1992.Cyclin E/cdk2 and cyclin A/cdk2 kinases associate with p107and E2F in a temporally distinct manner. Genes Dev. 6:1874-1885.

36. Loeken, M. R., and J. Brady. 1989. The adenovirus EIIAenhancer. Analysis of regulatory sequences and changes inbinding activity of ATF and EIIF following adenovirus infec-tion. J. Biol. Chem. 264:6572-6579.

37. Morris, S., M. Valentine, D. Shapiro, J. Sublett, L. Deaven, W.Roberts, J. Foust, D. Cerrentti, and T. Look. 1991. Reassign-ment of the human CSF1 gene to chromosome lpl3-p2l. Blood78:2013-2020.

38. Mudryj, M., S. H. Devoto, S. W. Hiebert, T. Hunter, J. Pines,and J. R. Nevins. 1991. Cell cycle regulation of the E2Ftranscription factor involves an interaction with cyclin A. Cell65:1243-1253.

39. Murre, C., P. S. McCaw, H. Vaessin, M. Caudy, L. Y. Jan, Y. N.Jan, C. V. Cabrera, J. N. Buskin, S. D. Hauschka, A. B. Lassar,H. Weintraub, and D. Baltimore. 1989. Interactions betweenheterologous helix-loop-helix proteins generate complexes thatbind specifically to a common DNA sequence. Cell 58:537-544.

40. Nakabeppu, Y., K. Ryder, and D. Nathans. 1988. DNA bindingactivities of three murine Jun proteins: stimulation by Fos. Cell55:907-915.

41. Nevins, J. R. 1992. E2F; A link between the Rb Tumor suppres-sor protein and viral oncoproteins. Science 258:424-429.

42. Pinkel, D., J. Landegent, C. Collins, J. Fuscoe, R. Segraves, J.

Lucas, and J. Gray. 1988. Fluorescence in situ hybridizationwith human chromosome-specific libraries: detection of trisomy21 and translocations of chromosome 4. Proc. Natl. Acad. Sci.USA 85:9138-9142.

43. Rauscher, F. J. I., R. R. Cohen, T. Curran, T. J. Bos, P. K.Vogt, D. Bohman, R. Tjian, and B. R. Franza, Jr. 1988.Fos-associated protein p39 is the product of the jun protoonco-gene. Science 240:1010-1016.

44. Schwarz, J. K., S. H. Devoto, E. J. Smith, S. P. Chellappan, L.Jakoi, and J. R. Nevins. 1993. Interactions of the p107 and Rbproteins with E2F during the cell proliferation response. EMBOJ. 12:1013-1020.

45. Shan, B., X. Zhu, P.-L. Chen, T. Durfee, Y. Yang, D. Sharp, andW.-H. Lee. 1992. Molecular cloning of cellular genes encodingretinoblastoma-associated proteins: identification of a gene withproperties of the transcription factor E2F. Mol. Cell. Biol.12:5620-5631.

46. Shirodkar, S., M. Ewen, J. A. DeCaprio, D. Morgan, D. Living-ston, and T. Chittenden. 1992. The transcription factor E2Finteracts with the retinoblastoma product and a p107-cyclin Acomplex in a cell cycle-regulated manner. Cell 68:157-166.

47. Sleigh, M. J. 1986. A nonchromatographic assay for expressionof the chloramphenicol acetyltransferase gene in eukaryoticcells. Anal. Biochem. 156:251-256.

48. Smith, D. B., and K. S. Johnson. 1988. Single-step purification ofpolypeptides expressed in Escherichia coli as fusions withglutathione S-transferase. Gene 67:3140.

49. Sumegi, J., E. Uzvolgyi, and G. Klein. 1990. Expression of theRB gene under the control of MuLV-LTR suppresses tumori-genicity of WERI-Rb-27 retinoblastoma cells in immunodefec-tive mice. Cell Growth Differ. 1:247-250.

50. Takahashi, R., T. Hashimoto, H.-J. Xu, S.-X. Hu, T. Matsui, T.Miki, H. Bigo-Marshall, S. A. Aaronson, and W. F. Benedict.1991. The retinoblastoma gene functions as a growth and tumorsuppressor in human bladder carcinoma cells. Proc. Natl. Acad.Sci. USA 88:5257-5261.

51. Thalmeier, K., H. Synovzik, R. Mertz, E.-L. Winnacker, and M.Lipp. 1989. Nuclear factor E2F mediates basic transcription andtrans-activation by Ela of the human MYC promoter. GenesDev. 3:527-536.

52. Wang, P. N., H. To, W. H. Lee, and E. Y. H. P. Lee. 1993.Tumor suppressor activity of RB and p53 genes in human breastcarcinoma cells. Oncogene 8:279-288.

53. Weinberg, R. A. 1992. The retinoblastoma gene and geneproduct. Cancer Surv. 12:43-57.

54. Weintraub, S. J., C. A. Prater, and D. C. Dean. 1992. Retino-blastoma protein switches the E2F site from positive to negativeelement. Nature (London) 358:259-261.

55. Wood, W. M., M. Y. Kao, D. F. Gordon, and E. C. Ridgway.1989. Thyroid hormone regulates the mouse thyrotropin n-sub-unit gene promoter in transfected primary thyrotropes. J. Biol.Chem. 264:14840-14847.

56. Wu, C.-L., and E. Harlow. Personal communication.57. Yamasaki, L., and L. Cao. Personal communication.58. Zamanian, M., and N. B. La Thangue. 1993. Transcriptional

repression by the Rb-related protein p107. Mol. Biol. Cell4:389-396.

59. Zhu, L., S. van den Heuvel, K. Helin, A. Fattaey, M. Ewen, N.Dyson, and E. Harlow. 1993. Inhibition of cell proliferation byp107, a relative of the retinoblastoma protein. Genes Dev.7:1111-1125.

VOL. 13, 1993

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 01

Feb

ruar

y 20

22 b

y 61

.137

.125

.206

.